When_Computation_Hugs_Intelligence_Content-Aware_Data_Processing_for_Industrial_IoT -中兴通讯论文.pdf

吾亦可往

10页

1次

2024-04-22

免费下载

IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO. 3, JUNE 2018 1657

When Computation Hugs Intelligence:

Content-Aware Data Processing for Industrial IoT

Liang Zhou , Dan Wu, Jianxin Chen, and Zhenjiang Dong

Abstract—Data service has been considered as one the most

prominent characteristics for Industrial Internet of Things (IIoT).

This paper studies how to design an optimal computing manner

for a general IIoT system. On the theory end, we analyze the

relationship between the data processing and the energy con-

sumption through investigating the content correlation of the

captured data. Importantly, we derive an exact expression for

the performance of IIoT by combining computation with intel-

ligence. On the application end, we design an efﬁcient way to

obtain a threshold by approximating the performances of differ-

ent computing manners, and show how to apply it to practical

IIoT applications. We believe that the proposed computation

rules hold great signiﬁcance for the IIoT designer, that is, it

is better to use distributed computing manner when the content

correlation is high, otherwise, centralized computing manner is

better.

Index Terms—Computation intelligence, computing manner,

data processing, Industrial Internet of Things (IIoT).

I. INTRODUCTION

N RECENT years, as the advances of sensor hardware,

computation capacity, and communications technology,

Industrial Internet of Things (IIoT), as a promising tool

and platform for Industry 4.0, has been widely studied

and employed in various scenarios [1]–[4]. Essentially, IIoT

deploys an integrated information technology to collect infor-

mation from different kinds of sensors, transmit it to the data

centers, and update the related parameters in the form of a

closed loop system [5]–[7].

From the view of the IIoT’s functions, it is clear that

data is placed as the core position [8]–[12]. Data collection,

Manuscript received September 15, 2017; revised December 2, 2017;

accepted December 15, 2017. Date of publication December 20, 2017; date of

current version June 8, 2018. This work was supported in part by the National

Natural Science Foundation of China under Grant 61571240 and Grant

61671474, in part by the Jiangsu Science Fund for Excellent Young Scholars

under Grant BK20170089, in part by the ZTE Program “The Prediction of

Wireline Network Malfunction and Trafﬁc Based on Big Data,” and in part

by the Priority Academic Program Development of Jiangsu Higher Education

Institutions. (Corresponding author: Liang Zhou.)

L. Zhou and J. Chen are with the National Engineering Research

Center for Communication and Network Technology and the College

of Communication and Information Engineering, Nanjing University of

Posts and Telecommunications, Nanjing 210003, China, and also with

the Jiangsu High Technology Research Key Laboratory for Wireless

Sensor Networks, Nanjing 210003, China (e-mail: liang.zhou@njupt.edu.cn;

chenjx@njupt.edu.cn).

D. Wu is with the Institute of Communications Engineering, Army

Engineering University of PLA, Nanjing 210007, China (e-mail:

wujing1958725@126.com).

Z. Dong is with the Cloud and IT Institute, ZTE Corporation,

Nanjing 210012, China (e-mail: dong.zhenjiang@zte.com.cn).

Digital Object Identiﬁer 10.1109/JIOT.2017.2785624

transmission, and application form the main parts of the IIoT,

in particular, data processing is the premise and it runs through

each part of IIoT [10], [13]. In current IIoT data processing

system, there are two main computing manners: 1) distributed

computing (DC) and 2) centralized computing (CC), in which

DC denotes that each data is processed by each sensor while

CC indicates each data is transmitted to the data center which

takes charge of the data processing [14]–[17].

Although these two computing manners have been widely

used for data processing in IIoT, it is still not clear how to

apply them for a speciﬁc application, e.g., for an IIoT designer,

which computing manner to choose? Why? Intuitively, DC and

CC both have distinct advantages and disadvantages. When the

obtained data are processed (e.g., compression, cleaning, etc.)

by each sensor, thus the transmitted data from the sensors to

the data centers can be reduced dramatically. As expected,

the energy consumption of the data communications can be

reduced greatly as well, but at the cost of the additional energy

consumption by the data processing at each sensor. As a result,

the optimal computing manner depends on achieving the min-

imal total energy consumption by balancing the processing

part and transmission part. We start by providing a practical

example.

A. Motivations

Fig. 1 shows the power consumption of the transportation

video and environmental monitoring services of Nanjing City,

China, which can be viewed as an IIoT for smart city. Given

these results, we have the following basic observations. For

the video services, the performance of DC is better than that

of CC, while CC has the advantage for the environmental

monitoring service. Why different computing manners have

so distinct performances for different applications? Here, a

nature and fundamental question rises: what is the difference

between DC and CC? Alternatively, we can ask even more

complicated questions: how do the DC and CC impact on

the performance of IIoT? How do these applications impact

the computing manner? How to apply the results to practi-

cal industrious applications? These questions form the main

motivations of this paper.

B. Contributions

In this part, we present an overview of the main contri-

butions. Exact interpretations and proofs of these results will

be provided in Section IV after the system model has been

introduced in Section III. This paper aims at studying the

2327-4662

 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.

See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: ZTE CORPORATION. Downloaded on August 19,2024 at 23:44:44 UTC from IEEE Xplore. Restrictions apply.

1658 IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO. 3, JUNE 2018

(a) (b)

Fig. 1. Consumed power of different services by different computing manners. The data of DC and CC are recorded in Nov. 2016 and Apr. 2017, respectively.

(a) Transportation video. (b) Environmental monitoring.

performance implications of different computing manners in

the environment of IIoT. In particular, we use the total energy

consumption E, which is considered as one of the most impor-

tant parameters of IIoT, as the performance index. Importantly,

we link the energy consumption E with the content relation-

ship of the transmitted data, as expressed by the correlation

coefﬁcient r. To this end, we propose a so-called relationship

estimation, whereby the content correlations of the transmitted

data at each sensor are estimated by a stochastic ﬂuid model.

In this paper, we systemically and fundamentally answer the

question of how to design an optimal computing manner for a

general IIoT. In particular, we design an efﬁcient and simple

content-aware data processing rule: it is better to use DC when

the content correlation is high, otherwise, CC is better. The

detailed contributions can be summarized as follows.

1) On the theory end, we analyze the relationship between

the transmitted data and the energy consumption which

results from the data processing and transmission.

Importantly, we derive an exact expression for the

stochastic ﬂuid model, for any transmitted data correla-

tion coefﬁcient r and computing manners (DC or CC),

thus characterizing the performance of the IIoT in terms

of E. Importantly, all the analysis is based on a general

computing framework, enabling these theoretical results

can be utilized to various IIoT systems and applications.

2) On the application end, we show that for any transmit-

ted data, the evolution of relation estimation process can

be approximated by a Markov process with a ﬁnite win-

dow. By setting the size of the ﬁnite window, we develop

a simple content analysis method to quickly estimate r

with high accuracy. Then, we design an efﬁcient way

to get a threshold by approximating the stochastic ﬂuid

model, and show how to apply the threshold to any prac-

tical IIoT applications. Numerical practical results have

demonstrated the effectiveness of the proposed method.

The rest of this paper is organized as follows. Section II pro-

vides an overview of the related works. Section III describes

the system model considered in this paper. In Section IV,

we proposes a content-aware data processing scheme based

on intelligently analyzing the content correlation of the

transmitted data. Subsequently, Section V provides numerical

results to validate the proposed system. Section VI concludes

this paper with a summary.

II. R

ELATED WORK

Generally speaking, the study of computing manner in the

context of IIoT is still in its infancy. Existing works usually

do not consider the impact of the transmission content on the

IIoT performance, and how to link the transmission content

with the computing manner and energy consumption is an

open problem for IIoT.

In a broad sense, the general problem of computing man-

ner on the system performance has been investigated widely.

Speciﬁcally, Bello and Zeadally [18] studied how much intel-

ligence that D2D communication can be achieved in the

IoT ecosystem. Tang et al. [19] proposed two opportunis-

tic random access mechanisms, i.e., overlapped contention

and segmented contention, to intelligently choose the best

channel condition. Moreover, Chen et al. [20] examined

the typical computing manner of narrow band IoT and dis-

cuss the performance tradeoffs in existing designs. Moreover,

Lai et al. [21] presented computation-based device ser-

vice approach that provides multimedia data suitable for a

terminal unit environment via interactive mobile streaming

services.

Theoretically, due to the heterogeneity of the network topol-

ogy, transmission content, processing capacity, and service

request, data processing in IIoT is a classical multiparame-

ter multiconstraint multicondition problem, and the traditional

convex optimization tools cannot be applied directly [22].

Instead, in this paper, we tactfully decompose this complex

progress into a series of stochastic optimization problems that

enable us to observe and investigate the performance of IIoT

in a deterministic manner.

Importantly, we extend existing works on two critical

aspects. First, we formulate the distributed/centralized com-

puting problem based on the real content correlation of the

observed data, and analyze the impacts of the computing man-

ner and transmission fashion on the total energy consumption

Authorized licensed use limited to: ZTE CORPORATION. Downloaded on August 19,2024 at 23:44:44 UTC from IEEE Xplore. Restrictions apply.

of 10

免费下载

goldendb paper

文档被以下合辑收录

GoldenDB论文（共12篇）

GoldenDB论文

关注

文档被以下合辑收录

评论